[YARN-2065] AM cannot create new containers after restart-NM token from previous attempt used

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.4.0
• Fix Version/s: 2.5.0
• Target Version/s: 2.5.0
• Assignee: Jian He
• Reporter: Steve Loughran
• Created: 15/May/14 11:47
• Updated: 15/Jul/16 09:11
• Resolved: 02/Jul/14 18:35

Description

Slider AM Restart failing (SLIDER-34). The AM comes back up, but it cannot create new containers.

The Slider minicluster test TestKilledAM can replicate this reliably -it kills the AM, then kills a container while the AM is down, which triggers a reallocation of a container, leading to this failure.

Issue Links

• duplicates: YARN-2371 Wrong NMToken is issued when NM preserving restarts with containers running
• duplicates: YARN-2433 Stale token used by restarted AM (with previous containers retained) to request new container
• is depended upon by: SLIDER-66 AM Restart not working -YARN issues
• relates to: SLIDER-34 Restarted AM cannot create containers
• relates to: FLINK-4142 Recovery problem in HA on Hadoop Yarn 2.4.1